1. Field of the Invention
The present invention relates to RF amplifiers and signal modulation.
2. State of the Art
Battery life is a significant concern in wireless communications devices such as cellular telephones, pagers, wireless modems, etc. Radio-frequency transmission, especially, consumes considerable power. A contributing factor to such power consumption is inefficient power amplifier operation. A typical RF power amplifier for wireless communications operates with only about 10% efficiency. Clearly, a low-cost technique for significantly boosting amplifier efficiency would satisfy an acute need.
Furthermore, most modern digital wireless communications devices operate on a packet basis. That is, the transmitted information is sent in a series of one or more short bursts, where the transmitter is active only during the burst times and inactive at all other times. It is therefore also desirable that control of burst activation and deactivation be controlled in an energy-efficient manner, further contributing to extended battery life.
Power amplifiers are classified into different groups: Class A, Class B, Class AB, etc. The different classes of power amplifiers usually signify different biasing conditions. In designing an RF power amplifier, there is usually a trade-off between linearity and efficiency. The different classes of amplifier operation offer designers ways to balance these two parameters.
Generally speaking, power amplifiers are divided into two different categories, linear and non-linear. Linear amplifiers (e.g. Class A amplifiers and Class B push-pull amplifiers), maintain high linearity, resulting in faithful reproduction of the input signal at their output since the output signal is linearly proportional to the input signal. In non-linear amplifiers (e.g. single-ended Class B and Class C amplifiers), the output signal is not directly proportional to the input signal. The resulting amplitude distortion on the output signal makes these amplifiers most applicable to signals without any amplitude modulation, which are also known as constant-envelope signals.
Amplifier output efficiency is defined as the ratio between the RF output power and the input (DC) power. A major source of power amplifier inefficiency is power dissipated in the transistor. A Class A amplifier is inefficient since current flows continuously through the device. Conventionally, efficiency is improved by trading-off linearity for increased efficiency. In Class B amplifiers, for example, biasing conditions are chosen such that the output signal is cut off during half of the cycle unless the opposing half is provided by a second transistor (push-pull). As a result, the waveform will be less linear. The output waveform may still be made sinusoidal using a tank circuit or other filter to filter out higher and lower frequency components.
Class C amplifiers conduct during less than 50% of the cycle, in order to further increase efficiency; i.e., if the output current conduction angle is less than 180 degrees, the amplifier is referred to as Class C. This mode of operation can have a greater efficiency than Class A or Class B, but it typically creates more distortion than Class A or Class B amplifiers. In the case of a Class C amplifier, there is still some change in output amplitude when the input amplitude is varied. This is because the Class C amplifier operates as a constant current sourcexe2x80x94albeit one that is only on brieflyxe2x80x94and not a switch.
The remaining classes of amplifiers vigorously attack the problem of power dissipation within the transistor, using the transistor merely as a switch. The underlying principle of such amplifiers is that a switch ideally dissipates no power, for there is either zero voltage across it or zero current through it. Since the switch""s V-I product is therefore always zero, there is no dissipation in this device. A Class E power amplifier uses a single transistor, in contrast with a Class D power amplifier, which uses two transistors
In real life, however, switches are not ideal. (Switches have turn on/off time and on-resistance.) The associated dissipation degrades efficiency. The prior art has therefore sought for ways to modify so-called xe2x80x9cswitch-modexe2x80x9d amplifiers (in which the transistor is driven to act as a switch at the operating frequency to minimize the power dissipated while the transistor is conducting current) so that the switch voltage is zero for a non-zero interval of time about the instant of switching, thereby decreasing power dissipation. The Class E amplifier uses a reactive output network that provides enough degrees of freedom to shape the switch voltage to have both zero value and zero slope at switch turn-on, thus reducing switching losses. Class F amplifiers are still a further class of switch-mode amplifiers. Class F amplifiers generate a more square output waveform as compared to the usual sinewave. This xe2x80x9csquaring-upxe2x80x9d of the output waveform is achieved by encouraging the generation of odd-order harmonics (i.e., xc3x973, xc3x975, xc3x977, etc.) and suppressing the even-order harmonics (i.e., xc3x972, xc3x974, etc.) in the output network.
An example of a known power amplifier for use in a cellular telephone is shown in FIG. 1. GSM cellular telephones, for example, must be capable of programming output power over a 30 dBm range. In addition, the transmitter turn-on and turn-off profiles must be accurately controlled to prevent spurious emissions. Power is controlled directly by the DSP (digital signal processor) of the cellular telephone, via a DAC (digital to analog converter). In the circuit of FIG. 1, a signal GCTL drives the gate of an external AGC amplifier that controls the RF level to the power amplifier. A portion of the output is fed back, via a directional coupler, for closed-loop operation. The amplifier in FIG. 1 is not a switch-mode amplifier. Rather, the amplifier is at best a Class AB amplifier driven into saturation, and hence demonstrates relatively poor efficiency.
FIG. 2 shows an example of a known Class E power amplifier, described in U.S. Pat. No. 3,919,656. An RF input signal is coupled over a lead 1 to a driver stage 2, the latter controlling the active device 5 via a signal coupled over a lead 3. The active device 5 acts substantially as a switch when appropriately driven by the driver 2. The output port of the active device is therefore represented as a single-pole single-throw switch 6. Connected across the switch 6 is the series combination of a DC power supply 7 and the input port of a load network 9. The output port of the load network 9 is connected to the load 11. As the switch 6 is cyclically operated at the desired AC output frequency, DC energy from the power supply 7 is converted into AC energy at the switching frequency (and harmonics thereof).
U.S. Pat. No. 3,900,823 to Sokal et al. describes feedback control of Class E power amplifiers. The need for feedback control suggests the inability to fully characterize device behavior, which in turn suggests substantial departure from operation of the device as a true switch. Sokal further describes a solution to the problem of feedthrough power control at low power levels by controlling RF input drive magnitude through application of negative feedback techniques to control the DC power supply of one or more preceding stages. The need for feedback control imposes constraints of feedback loop dynamics on a system.
The Class E amplifier arrangement of FIG. 2, although it is theoretically capable of achieving high conversion efficiency, suffers from the disadvantage that large voltage swings occur at the output of the active device, due to ringing. This large voltage swing, which typically exceeds three times the supply voltage, precludes the use of the Class E circuit with certain active devices which have a low breakdown voltage.
To operate an RF power amplifier in switch mode, it is necessary to drive the output transistor(s) rapidly between cutoff and full-on, and then back to cutoff, in a repetitive manner. The means required to achieve this fast switching is dependent on the type of transistor chosen to be used as the switch: for a field-effect transistor (FET), the controlling parameter is the gate-source voltage, and for a bipolar transistor (BJT, HBT) the controlling parameter is the base-emitter current.
However, the driving circuit in the RF amplifier of FIG. 2 typically includes a matching network consisting of a tuned (resonant) circuit. Referring to FIG. 3, in such an arrangement, an RF input signal is coupled to a driver amplifier, typically of Class A operation. An output signal of the driver amplifier is coupled through the matching network to a control terminal of the switching transistor, shown in FIG. 3 as an FET. As with design of the load network of FIG. 2, proper design of the matching network is not an easy matter.
Various designs have attempted to improve on different aspects of the basic Class E amplifier. One such design is described in Choi et al., A Physically Based Analytic Model of FET Class-E Power Amplifiersxe2x80x94Designing for Maximum PAE, IEEE Transactions on Microwave Theory and Techniques, Vol. 47, No. 9, September 1999. This contribution models various non-idealities of the FET switch and from such a model derives conclusions about advantageous Class E amplifier design. For the chosen topology, maximum power-added efficiency (PAE) of about 55% occurs at a power level of one-half watt or less. At higher powers, PAE is dramatically reduced, e.g., less than 30% at 2W.
The PAE of a power amplifier is set by the amount of DC supply power required to realize the last 26 dB of gain required to achieve the final output power. (At this level of gain, the power input to the amplifier through the driving signalxe2x80x94which is not readily susceptible to measurementxe2x80x94becomes negligible.) Presently, there are no known amplifying devices capable of producing output powers of 1W or greater at radio frequencies and that also provide a power gain of at least 26 dB. Accordingly, one or more amplifiers must be provided ahead of the final stage, and the DC power consumed by such amplifiers must be included in the determination of overall PAE.
Conventional design practice calls for an amplifier designer to impedance-match the driver output impedance to the input impedance of the final switching transistor. The actual output power therefore required from the driver stage is defined by the required voltage (or current) operating into the (usually low) effective input impedance of the switching element. A specific impedance for the input of the switching transistor is not definable, since the concept of impedance requires linear operation, and a switch is very nonlinear.
An example of an RF amplifier circuit in accordance with the foregoing approach is shown in FIG. 4. An interstage xe2x80x9cT sectionxe2x80x9d consisting of an inductor L1, a shunt capacitor C and an inductor L2 is used to match the driver stage to an assumed 50 ohm load (i.e., the final stage).
This conventional practice treats the interstage between the drive and final stages as a linear network, which it is not. Further, the conventional practice maximizes power transfer between the driver and final stages (an intended consequence of impedance matching). Thus, for example, in order to develop the required drive voltage for a FET as the switching transistor, the driver must also develop in-phase current as well to provide the impedance-matched power.
Another example of a conventional RF power amplifier circuit is shown in FIG. 5. This circuit uses xe2x80x9cresonant interstage matchingxe2x80x9d in which the drive and final stages are coupled using a coupling capacitor Ccpl.
As noted, conventional design practice fails to achieve high PAE at high output power (e.g., 2W, a power level commonly encountered during the operation of a cellular telephone). A need therefore exists for an RF power amplifier that exhibits high PAE at relatively high output powers.
Survey of Prior Patents
Control of the output power from an amplifier is consistently shown as requiring a feedback structure, as exemplified by Sokal et al. and further exemlified in the following U.S. Pat. Nos. 4,392,245; 4,992,753; 5,095,542; 5,193,223; 5,369,789; 5,410,272; 5,697,072 and 5,697,074. Other references, such as U.S. Pat. No. 5,276,912, teach the control of amplifier output power by changing the amplifier load circuit.
A related problem is the generation of modulated signals, e.g., amplitude modulated (AM) signals, quadrature amplitude modulated signals (QAM), etc. A known IQ modulation structure is shown in FIG. 6. A data signal is applied to a quadrature modulation encoder that produces I and Q signals. The I and Q signals are applied to a quadrature modulator along with a carrier signal. The carrier signal is generated by a carrier generation block to which a tuning signal is applied.
Typically, an output signal of the quadrature modulator is then applied to a variable attenuator controlled in accordance with a power control signal. In other instances, power control is implemented by vaying the gain of the amplifier. This is achieved by adjusting the bias on transistors within the inear amplifier, taking advantage of the effect where transistor transconductance varies with the aplied bias conditions. Since amplifier gain is strongly related to the transistor transconductance, varying the transconductance effectively varies the amplifier gain. A resulting signal is then amplified by a linear power amplifier and applied to an antenna.
In AM signals, the amplitude of the signal is made substantially proportional to the magnitude of an information signal, such as voice. Information signals such as voice are not constant in nature, and so the resulting AM signals are continuously varying in output power.
A method for producing accurate amplitude modulated signals using non-linear Class C amplifiers, called xe2x80x9cplate modulation,xe2x80x9d has been known for over 70 years as described in texts such as Terman""s Radio Engineers Handbook (McGraw-Hill, 1943). In the typical plate-modulation technique, output current from the modulator amplifier is linearly added to the power supply current to the amplifying element (vacuum tube or transistor), such that the power supply current is increased and decreased from its average value in accordance with the amplitude modulation. This varying current causes the apparent power supply voltage on the amplifying element to vary, in accordance with the resistance (or conductance) characteristics of the amplifying element.
By using this direct control of output power, AM can be effected as long as the bandwidth of the varying operating voltage is sufficient. That is, these non-linear amplifiers actually act as linear amplifiers with respect to the amplifier operating voltage. To the extent that this operating voltage can be varied with time while driving the nonlinear power amplifier, the output signal will be linearly amplitude modulated.
Other methods of achieving amplitude modulation include the combination of a multitude of constant amplitude signals, as shown in the following U.S. Pat. Nos. 4,580,111; 4,804,931; 5,268,658 and 5,652,546. Amplitude modulation by using pulse-width modulation to vary the power supply of the power amplifier is shown in the following U.S. Pat. Nos. 4,896,372; 3,506,920; 3,588,744 and 3,413,570. However, the foregoing patents teach that the operating frequency of the switch-mode DC-DC converter must be significantly higher than the maximum modulation frequency.
U.S. Pat. No. 5,126,688 to Nakanishi et al. addresses the control of linear amplifiers using feedback control to set the actual amplifier output power, combined with periodic adjustment of the power amplifier operating voltage to improve the operating efficiency of the power amplifier. The primary drawback of this technique is the requirement for an additional control circuit to sense the desired output power, to decide whether (or not) the power amplifier operating voltage should be changed to improve efficiency, and to effect any change if so decided. This additional control circuitry increases amplifier complexity and draws additional power beyond that of the amplifier itself, which directly reduces overall efficiency.
A further challenge has been to generate a high-power RF signal having desired modulation characteristics. This object is achieved in accordance with the teachings of U.S. Pat. No. 4,580,111 to Swanson by using a multitude of high efficiency amplifiers providing a fixed output power, which are enabled in sequence such that the desired total combined output power is a multiple of this fixed individual amplifier power. In this scheme, the smallest change in overall output power is essentially equal to the power of each of the multitude of high efficiency amplfiers. If finely graded output power resolution is required, then potentially a very large number of individual high efficiency amplifiers may be required. This clearly increases the overall complexity of the amplifier.
U.S. Pat. No. 5,321,799 performs polar modulation, but is restricted to full-response data signals and is not useful with high power, high-efficiency amplifiers. The patent teaches that amplitude variations on the modulated signal are applied through a digital multiplier following phase modulation and signal generation stages. The final analog signal is then developed using a digital-to-analog converter. As stated in the State of the Art section herein, signals with information already implemented in amplitude variations are not compatible with high-efficiency, nonlinear power amplifiers due to the possibly severe distortion of the signal amplitude variations.
Despite the teachings of the foregoing references, a number of problems remain to be solved, including the following: to achieve high-efficiency amplitude modulation of an RF signal by varation of the operating voltage using a switch mode converter without requiring high-frequency switch-mode operation (as compared to the modulation frequency); to unify power-level and burst control with modulation control; to enable high-efficiency modulation of any desired character (amplitude and/or phase); and to enable high-power operation (e.g., for base stations) without sacrificing power efficiency.
The present invention, generally speaking, provides for high-efficiency power control of a high-efficiency (e.g., hard-limiting or switch-mode) power amplifier in such a manner as to achieve a desired control or modulation. Unlike the prior art, feedback is not required. That is, the amplifier may be controlled without continuous or frequent feedback adjustment. In one embodiment, the spread between a maximum frequency of the desired modulation and the operating frequency of a switch-mode DC-DC converter is reduced by following the switch-mode converter with an active linear regulator. The linear regulator is designed so as to control the operating voltage of the power amplifier with sufficient band-width to faithfully reproduce the desired amplitude modulation waveform. The linear regulator is further designed to reject variations on its input voltage even while the output voltage is changed in response to an applied control signal. This rejection will occur even though the variations on the input voltage are of commensurate or even lower frequency than that of the controlled output variation. Amplitude modulation may be achieved by directly or effectively varying the operating voltage on the power amplifier while simultaneously achieving high efficiency in the conversion of primary DC power to the amplitude modulated output signal. High efficiency is enhanced by allowing the switch-mode DC-to-DC converter to also vary its output voltage such that the voltage drop across the linear regulator is kept at a low and relatively constant level. Time-division multiple access (TDMA) bursting capability may be combined with efficient amplitude modulation, with control of these functions being combined. In addition, the variation of average output power level in accordance with commands from a communications system may also be combined within the same structure.
The high-efficiency amplitude modulation structure may be extended to any arbitrary modulation. Modulation is performed in polar form, i.e., in a quadrature-free manner.
Single high-efficiency stages may be combined together to form high-power, high-efficiency modulation structures.